Maven Repository Managers - why we chose Archiva...and then switched to Nexus

It looks like the Maven folk have updated the Archiva web site home page (http://maven.apache.org/archiva/), possibly to correspond with the recent 1.0.1 release. Nice work, guys, let's hope we see more of it for the other products!

When it came to choosing a Maven repository manager, Archiva was our choice. This article explains why. 

For anyone who doesn't know, Archiva is a Maven repository manager. If you are using Maven in a corporate environment, then you probably need a repository manager. A Maven repository manager acts as both a repository for your JAR files, and a proxy/cache between you and all those dodgy public repositories out there on the internet. So, when Maven asks for a new dependency, its first port of call is your local repository. If the JAR has already been downloaded, it will be stored in the Archiva repository, and you won't need to download it from the internet. This is both faster and more reliable than downloading JAR files directly from the internet.

You can also use a repository/proxy to limit the places developers can download JAR files from. This is useful if you want to make sure people stick to stable builds, and don't use the often-unstable SNAPSHOT builds for their libraries or plugins. Although SNAPSHOT builds may be tempting, they are in fact the bane of a stable build process, and should only be used for experimental purposes. Snapshot plugins? Just say no.

Anyway, so why did we choose Archiva? Archiva is easy to install and use, and has a decent feature set. You can manage user rights, access theHowever, the feature that made us prefer Archiva to the others is its configuration. You can configure Archiva either via an XML configuration file, or through the web interface. This includes adding new network proxies, which means you can add a new network proxy through the web interface without having to restart the server.

Now obviously, Archiva is not the only repository manager around. There are also a few others around. However, none of them met all of our requirements for a repository manager. Here's the run-down:

Artifactory is another Maven repository manager with proxying and caching features. The user interface is appealing. It also has nice Webdav support, and it is easy to import an existing file-based repository into the Artifactory repository. However, you can't configure the server via the web interface - if you want to add a new remote repository, you need to modify a (albeit easy-to-understand) XML configuration file and restart the server.

Proximity is the other major repository manager on the market. It is quite powerful, but I found it a bit complicated to configure: to add a new remote repository, you need to update an XML configuration file and a properties file, and restart the server.

Our choice of a Maven repository manager needed to take into account several factors, including:

• Can you manage the configuration via the web interface, without restarting the server?
• Is the user interface easy to understand?
• Is it easy to use from Maven
• Can the product host local repositories (we need to be able to store and share internal JARs)
• Can the product act as a proxy for remote repositories
• Does the product caches downloaded files?
• Is it easy to backup and restore the repository automatically?

As it turns out, Archiva is the only product that supports full web-based configuration. I found this a little suprising, since you would think that a repository proxy would benefit from being highly available, and if it has to go down every time you add a new external repository, that might have some impact on availability...

Archiva also supports the other required features in our list. For a production environment, considerations such as modifying the configuration without having to restart the server, and being able to backup the repository as a file system, are particularly important. Automatically  backing up Archiva simply requires backing up the data directory, which contains the cached repositories (stored as an ordinary file structure) and the databases that Archiva uses for its own purposes. Archiva uses two separate databases: one for the users, and one to basically index the repository files. You really need to back up the user database. The other database isn't that critical, since it will be regenerated from scratch if required, Another solution, more appropriate for a production environment, is to set up two external database connections for Archiva's two databases, and then to back up the cached repository file structure separately.

Now for the down side. Any product has its drawbacks, and Archiva is no exception. The user interface is initially slow when the JSP pages compile, but once that is done, it's fast enough. A bit clunky, but fast enough.
Archiva's biggest drawback comes from it's biggest advantage - the web configuration interface is powerful and flexible, but it takes some time to get your head around how it works. Nevertheless, at the end of the day, Archiva does do the job, and does it well.

For more on this subject, Deng Ching, one of the Archiva developers, discusses Archiva and the other repository managers in some detail here.